Here’s where the AI arms race stands after a week of hectic hype

Do you hear that? That’s the sound of the tech press finally getting a chance to take a breath.

OpenAI and Google unloaded a barrage of announcements this week, leading some enthusiasts to proclaim that the dystopian future from the 2013 movie Her had arrived, while more skeptical observers questioned where the AI hype has gotten us after months of breathless rollouts.

In any case, the week showed that competition between tech giants trying to own the generative AI space is still fierce, as the companies seemed determined to one-up one another in a race toward realizing multimodal all-purpose assistants.

We’ll get to the state of that arms race, but first, a quick recap of the news:

• No sign yet of GPT-5 or a rumored search competitor, but OpenAI announced on Monday a new model called GPT-4o (the “o” stands for “omni”) that will be widely available as part of ChatGPT’s free version. The company claimed the model is faster and cheaper than previous versions, with an improved understanding of vision and audio inputs.
• OpenAI employees wowed the audience with some on-stage demos of an intonation-heavy—and, as some observed, seductive—female voice for the new model interacting with images and performing translations. CEO Sam Altman claimed this voice mode would be available in “coming weeks.”
• Two OpenAI leaders—chief scientist Ilya Sutskever and alignment researcher Jan Leike—announced their departures from the company.
• Gemini Nano, Google’s phone-based foundation model, will also see more multimodal capabilities, meaning it will be able to better understand vision, audio, and voice, according to the company.
• Google announced an expansion of AI-generated summaries in search results, likely unnerving publishers who rely on its search traffic.
• Google showed off a “vision for the future of AI assistants” called Project Astra, an updated video model, and some other projects in varying degrees of completion.

Multimodal mania: While both companies are trying myriad AI-related things right now, there were some key themes in the news flurry. Both companies are clearly still chasing a vision of a multimodal assistant that can flit between text, voice, and other media with ease.

Chirag Dekate, VP analyst at Gartner Research, said the mix of concrete announcements with further-afield visions in both sets of announcements reflects an effort to strike a balance between the demands of competition and a fear of backlash around half-baked releases.

“The frantic pace of activity is now guardrailed internally and externally with the controls in place,” Dekate said. “And I think the companies involved are doing their darnedest to minimize any embarrassing headlines in the future.”

Point, Google? Dekate said he has been impressed with Google’s recent ability to “execute like an AI-native company” and its 2-million-token context window—an expanded ability to digest large tracts of information—on certain models, which could be useful in certain enterprise contexts.

“Google is now setting the basis of where the market needs to go,” Dekate said. “Suddenly, OpenAI is playing catch-up. And I think this capability gap is going to widen. That’s a very strong statement…The reason I have this hypothesis is because Google has now activated all of its engines.”

Microsoft, meanwhile, is more reliant on partners like OpenAI and Mistral for innovation, Dekate said. But Amazon, which is building on an established voice chatbot while OpenAI is working toward one, could also be a surprise contender, according to Dekate.

“Amazon is a dark horse—nobody’s paying attention to Amazon, but they have something cooking that is quite special,” Dekate said. “They have the cloud scale to deliver it very quickly to the market.”